Nonsteroidal anti-inflammatory drugs (NSAIDs) are widely used for the relief of the symptoms of osteoarthritis (OA), rheumatoid arthritis (RA), sprains and strains, sports injuries and menstrual disorders, and have a small role in the management of patent ductus arteriosus in the neonate. In patients with RA, symptom relief through use of NSAIDs is firmly established, although it remains unclear whether they influence the course and outcome of the disease. For the average patient with RA taking NSAIDs, the attributable risk of hospitalisation with gastrointestinal problems related to NSAIDs is 1.3 to 1.6% annually and risk of death is 0.15%. Associations of therapy with risk are greatest with age, corticosteroid use and previous NSAID-related gastrointestinal adverse effects, and less marked with disability and high NSAID dose. These are important data in attempting to balance risk of therapy with clinical efficacy in an individual patient, and assessing the cost-effectiveness of prophylaxis. Although half of all NSAID consumption is for control of pain associated with degenerative conditions, their superiority over simple analgesics in osteoarthritis is poorly documented. This finding supports the use of the simple analgesic paracetamol (acetaminophen) as the preferred therapy of osteoarthritis, especially when its lower cost and low incidence of adverse effects are taken into consideration. Consistent differences in clinical effectiveness of individual NSAIDs have not been demonstrated, although unpredictable interpatient variation in response to individual agents is of considerable clinical importance, and a more expensive NSAID may prove cost effective for some patients. Cost effectiveness can be improved by a self-adjusted dosage regime which also leads to lower overall drug consumption. The adverse gastrointestinal effects of these drugs account for about 30% of the overall cost of arthritis treatment, and although studies to date have been too limited to assess the relative risk of gastrointestinal toxicity of the different NSAIDs reliably, ibuprofen appears to be one of the least hazardous, and azapropazone one of the most hazardous. Although the effectiveness of prophylaxis with H 2-antagonists and with prostaglandin E 1 analogues (prostaglandin-E 1 analogues) has been established, estimates of cost-benefit ratios are widely divergent. To establish the most cost-effective therapy with NSAIDs, more data are required to establish multivariable risk profiles for identification of patients at particular risk, the optimal drug, and its optimal dosage and duration of treatment.(ABSTRACT TRUNCATED AT 400 WORDS)